Efficiently join same data twice without CTE in MySQL? - mysql

Question

I have this query that sums payments and reimbursements against pledges. It works, but smells bad:
select P.pledgeID, P.decamount,
(
sum(coalesce(C1.decamount, 0)) - sum(coalesce(C2.decamount, 0))
) as paymentTotal
from Pledge P
left join (select C.*, CT.eaddOrSubtract
from `Payment` C
left join PaymentType CT on C.paymentTypeID = CT.paymentTypeID )
C1 on P.pledgeID = C1.pledgeID and C1.eaddOrSubtract = 'add'
left join (select C.*, CT.eaddOrSubtract
from `Payment` C
left join PaymentType CT on C.paymentTypeID = CT.paymentTypeID)
C2 on P.pledgeID = C2.pledgeID and C2.eaddOrSubtract = 'subtract'
group by pledgeID
Particularly, I think there should be a better way to handle the joins inside the joins, especially since they produce the same results. On another RDBMS, I'd use a CTE, but that's not available here. Is there a more efficient way to calculate these payment totals (taking into account the fact that some are net additions and other net subtractions)?
Schema info:
PaymentType
---
| paymentTypeID | eaddOrSubtract | ...
| 1 | add |
| 2 | add |
| 3 | subtract |
| 4 | add |
| 5 | subtract |
Payment
---
| checkID | pledgeID | paymentTypeID | decamount | ...
| 1 | 19415 | 4 | 15.19 |
| 2 | 19414 | 2 | 900.00 |
| 3 | 19106 | 5 | 3856.00 |
| 4 | 19106 | 3 | 52.00 |
| 5 | 19414 | 1 | 15.00 |

Answer 1

The query should select all pledges (their pledgeID and decamount) and the total of payments for each pledge. Some payments are positive, some negative.
You query selects all pledges, joins the positive payments to each pledge and joins the negative payments to each row related to the pledge. If there is at most one negative and at most one positive payment it almost works (except that it returns NULL instead of 0 when there are no payments). Once there are at least two payments in one category (positive/negative) and at least one in the other category, problems arise. Each negative payment is joined to each positive payment on the same pledge and all the pairs are summed.
A cleaner way to look at the problem directly follows the first paragraph of this answer. Instead of two joins with filter on eaddOrSubtract, it suffices to make one join to a subquery, where the subquery internally handles the sign of the amount being summed. The CASE operator is great for such a job.
SELECT
P.pledgeID
, P.decamount
, COALESCE(SUM(C.signedDecamount), 0) AS paymentTotal
FROM Pledge P
LEFT JOIN (
SELECT
C.*
, CASE CT.eaddOrSubtract
WHEN 'add' THEN C.decamount
WHEN 'subtract' THEN -C.decamount
END AS signedDecamount
FROM Payment C
LEFT JOIN PaymentType CT ON C.paymentTypeID = CT.paymentTypeID
) C ON P.pledgeID = C.pledgeID
GROUP BY P.pledgeID
The COALESCE() call is there for the case when no payments are joined to the pledge or all they joined payments have NULL decamount. COALESCE() to 0 inside SUM() can always be safely omitted, as SUM() skips NULLs; I guess those calls were just artifacts of hacking the corner cases of joins in the original query.
SQL Fiddle